An idea that was born from the research areas of hypertext and information management.

If you consider a traditional text based thesaurus, it organises words into entries. Each entry represents a concept (much like an everything2 node) and lists synonyms for the concept. Given this structure, you can look up other synonyms for a concept, or find which concepts are described by a given word.

Text based information retrieval systems often include a thesaurus. This allows queries containing the word "car" to retrieve documents containing the word "automobile". The thesaurus enables automatic synonym substitution to be used.

A multimedia thesaurus is an attempt to adopt a traditional thesaurus approach but with information represented in several media. The concepts may be the same as a thesaurus, but the multimedia synonyms are very different.

For example, a multimedia thesaurus entry for the concept of "table" may have the word "table", a picture of a table, a sound clip of someone saying "table" and a 3D computer model of a table. During a query, any of these representations may be matched against to arrive at the concept, and any representation may be used as a result for the query.

Clearly, such an approach requires multimedia matching techniques to allow pictures and sounds and other representations to be compared. This is a whole research area in itself and is very much an unsolved problem in the general case. One problem is the way the same object can look very different when viewed from different angles. The computer can't match two different views of the same object easily.

A multimedia thesaurus can alleviate this problem by including several views as representations of a concept. This way, matching against any of the views will allow the same concept to be visited. It can be used as an approach to object classification.

The utility of the thesaurus can be enhanced by associating different information with the entries. If documents are associated with entries that represent the concepts they describe, then the thesaurus can be used as an index to the documents. Alternatively, relationships can be stored between the entries themselves, allowing the thesaurus to become a semantic network.